No, I do not know any company that could do that. And the posts since that agree with that. You are just trolling.
No, I do not know any company that could do that. And the posts since that agree with that. You are just trolling.You know who it is. Stop pretending that just because they're a lifestyle company they can't make good CPUs.
And go look at my original post. There is only one CPU series with a SPECint 1T score high enough and clock rate low enough to be nearly 50% ahead per GHz.
And from the patent trail from last quarter you'd imagine AMD is a packaging specialist than CPU designer. Not good. Typically patent trail is good indicator of active research in novel ideas but not since last two quarters unfortunately. At least not much applications in core design.
It said >10% as I recall which is code for 11-12%. But anyway, they need to over-deliver in the consumer space because they're behind.That slide could be red herring, there is no way that AMD only manages only 5-10% IPC in 2 years on N2 and if they do it’s because of Zen architecture needing a revamp and not cause x86 is dead.
Why on earth would PC CPUs hit some performance ceiling and phone CPUs wouldn't? Also despite all the hype AMD/Intel delivered larger IPC jumps than Apple since M1 launch. M2/M3/M4 are mostly just higher clocked/higher power versions of Firestorm.PC has, despite all the hate over many years, continually increased performance. And so it had the highest available performance for over 20 years. Until it was finally passed in 2024. If - like Hulk is saying - 5% after 2 years is the best they can ship do you think it can still retain its relevance?
At least AMD's slide for Zen 6 was more optimistic than that.
Heeyyyy M3/M4 are actually different even if most (all) the perf is clk bumps.M3/M4 are mostly just higher clocked/higher power versions of Firestorm.
Yes, but I suspect that the lions share of the PROFIT came from DC. Additionally, DC is a market that is increasing rapidly while client? Not so much.What do you mean client workloads don't matter? Client was over 25% percent of AMD's 2024 revenue.
Perhaps for the 32c variant (Zen 6c) on DC and possibly on client. Otherwise, I fully expect AMD to go with N3P for its desktop and laptop parts. This is necessary to maintain their margins and profit IMO.Zen 6 will likely use 2nm
The biggest difference between RDNA3 and RDNA4 is the return to a monolithic die IMO. The density gains are only partly the improved process. Not having a bunch of interconnect logic on a bunch of chiplets is likely the cause of the lower density on RDNA3.More avenues to scaling exist now, for performance. Also... On the topic of dnsity. Have a look at the latest RDNA4 density numbers vs RDNA3. Big gains still have been found. Though that's not the main point, which is that there are more ways to eek out performance.
Cost is going up, so have to be smarter about when to use bleeding edge and when not to. What will new materials bring. Glass substrates. Optical conputing. More to come...
Gapped in what? The strength of x86 is, has been, and always will be the incredible plethora of applications that are written for it. Seems like Apple has forgotten (as many here have .... or are too young to remember) how they nearly went bankrupt with their lack of app support. Used to go into the computer store to purchase programs. One tiny little shelf of Apple stuff, and an entire shop of x86 programs.Current x86 cores are just bad client cores, plain and simple. In what world it is acceptable for a laptop core to be gapped by a phone (the iPhone, to be precise) in running browsers.
But not profit.... and the trend is quite clear.And chances are, AMD client CPU revenue will surpass datacenter CPU revenue again this year.
Until a phone can do everything my laptop can, this is kind of a silly statement. Today, phones can't do even a small fraction of what a laptop does ..... and that doesn't even account for my previous x86 discussion on program compatibility.People, especially younger ones, already tend to use their phones as their main computing device. If phones start beating laptops in consumer workloads, then the whole segment will fade into irrelevance (DT already did).
There will be no need for x86.
How do you figure? The compute tile is already completely optimized for compute. I don't believe that companies (and certainly not AMD) will go back to monolithic in server, desktop and laptop segments. In fact, I think that even though RDNA4 went back, it is temporary. It simply doesn't make sense in MOST markets to pay the very high cost of a leading edge node and then only get a small number of good dies out of that very expensive wafer.As for Zen6, Nanoflex on N2 will remove the frequency stagnation. Fun times ahead.
No it's like all N2 besides lower-end mobile stuff (KRK replacements).Perhaps for the 32c variant (Zen 6c) on DC and possibly on client. Otherwise, I fully expect AMD to go with N3P for its desktop and laptop parts. This is necessary to maintain their margins and profit IMO.
Nope it's the microarchitecture.The biggest difference between RDNA3 and RDNA4 is the return to a monolithic die IMO
the opposite really, RDNA3 GCDs should be inherently higher density since they lack the MALL pile and all the shoreline required for DDR to work.Not having a bunch of interconnect logic on a bunch of chiplets is likely the cause of the lower density on RDNA3.
whatHow do you figure?
Interesting point. In fact, one could argue that Arrow Lakes failure is completely related to the poor inter-tile latencies introduced. AMD also suffered a setback with Zen 2, but is now somehow able to alleviate (or greatly so) the latency problems between die on a MCM. Intel needs to catch up rather quickly IMO. This seems to be where the large improvements seem to be lurking vs process density as in the past.It's funny that for a decade, Intel (with Foveros) was to be the "packaging specialist", but so far, Intel chiplet and packaging strategy / execution have been very much sub par.
I think the biggest change will be the IOD improvements. Faster memory, and lower latency may well account for the largest changes we see. I suspect that a few tweaks to the core will also be introduced, but those might be pretty minor in the overall scheme.Zen 6 may just be putting back in things that did not fit into Zen 5 (for faster turnaround) and mostly about new packaging / new modularity to cover the entire mid to premium to high end segments. Plus new process nodes. Plus, perhaps adding AMX instructions from Intel server CPUs to all AMD CPUs.
Zen 3 and M1 were on par. 5950X was 9.15 and M1 Pro was 9.2 in SPECint 2017 1T.Why on earth would PC CPUs hit some performance ceiling and phone CPUs wouldn't? Also despite all the hype AMD/Intel delivered larger IPC jumps than Apple since M1 launch. M2/M3/M4 are mostly just higher clocked/higher power versions of Firestorm.
no LNC is just pretty bad and their L3 is even moreso bad.In fact, one could argue that Arrow Lakes failure is completely related to the poor inter-tile latencies introduced
wonder what's the difference.Zen 5 and M4 are not on par. 9950X is 12.6 and M4 Pro is 13.6. Firestorm derivatives are pulling away.
The biggest change is the goddamn double-shrink.I think the biggest change will be the IOD improvements
M1 was node ahead and Zen 3 could still compete. It's not the real reason.Definitely not the node. Can't be that right?
Power went up M1 to M4 by a solid chunk.M1 was node ahead and Zen 3 could still compete
yeah it is.It's not the real reason.
Yep.The real difference is Apple is throwing power around now too.
The day before release as I recall.Will we see some 9950X3D reviews soon or will it not be until the 12th?
It's been a while since I've digested a good CPU review.
But not profit.... and the trend is quite clear.
you do understand that they can grow CPU and GPGPU share in DC tooAt 40% of the market, client segment would surpass server in both absolute revenue and also in operating margin.
you do understand that they can grow CPU and GPGPU share in DC too
no.GPGPU is in its own universe - as far as total size of the market and then market share.
34% is rev share, not units.But I think there is more potential to grow in client, since AMD is starting from lower base. IIRC, ~22% in CPU client and ~34% in CPU server.
34% is rev share, not units.
I know, right. Gates of Oblivion and all.It's going to suck up all of the resources of the Earth to continue to grow, like in the "Oblivion"
indeed.I am guessing that AMD unit share in datacenter CPU is lower than revenue share. Since AMD ASPs are most likely higher than Intel ASPs.
Saying that IPC is reduced by the operating system is like saying that my car's horsepower goes down when your mom gets in.
My mom drives slower than me. My car would get slower with my mom driving.
Casinos are not video games.
Versus what? The competition from Intel? I know people want to bring up Apple, but when Intel's overall marketshare in client is still higher than Apple's, it makes sense for AMD to gobble up as much of that as they can instead of worrying about Apple. Even if Zen6 only manages a paltry 10% increase in IPC, it's probably going to gain ground on Intel in client.Zen 6 still needs to be on the 19% side of 1x% to be a success in client.
20 years? Oh Kepler, it has been on death row since the 80s. 🤣"PC is dying" is a 20-year old trope and never been true. In fact latest research shows the largest growth market for things like gaming is PC.
Apple has stagnated. They have relied on adding things like SME and higher clocks in order to push performance up.M1 was node ahead and Zen 3 could still compete. It's not the real reason.
The real difference is Apple is throwing power around now too.
Zen 6 still needs to be on the 19% side of 1x% to be a success in client.
Intel has a yearly cadence. AMD will need Zen 6 simply to recover ground lost to Intel delivering new products. OEMs love saying they have something new. And since the mobile Zen 6 parts apparently aren't launching until 2027Q1 it might be two "generations" of losing ground to Intel.it's probably going to gain ground on Intel in client
Apple M4 is already ahead of Zen 6 mobile parts in SPECint 2017 1T unless AMD improves by 35%+ for laptop parts. Is anyone claiming Zen 6 will be that much of an improvement in these benchmarks? You want to start another hype train to run off a cliff? Zen 6 is designed to be a smaller core so they can fix 12 in 70-80mm. It's just the nature of Zen to be focused on server and that establishes an upper limit on size and a preference for a large L3 instead of L2. That Zen presently managed to be better than two different Intel cores shows what disarray Intel is in, not how great Zen 5 is.I am guessing another 20-25% total improvement for single core. We shall see.
Apple M4 is
Most of the best software are written on Macs and hosted on Linux servers. Windows is terrible for development nowadays.Irrelevant to most PC users.
I don’t even do software development on a Mac anymore.